I don't see anything wrong with the premise of the service or even from what I understand the technical execution.
Seems like the failure was around business development, marketing, community management, and brand reputation.
I haven't followed to closely though so other would know better than me.
I think the real failing is what you stated - biz dev and brand reputation.
Read the other replies and you will see folks say "There were no games" or "The selection was limited." Why in the world would a major game company hesitate to invest in google and google services? Well, if you have a reputation of killing off your services like the grim reaper, I imagine that would come up in an executive risk assessment.
Google can fix their reputation but its going to take a lot of work. Consider Microsoft's reputation in the late mid to late 90's on slashdot and other forums compared to their reputation today. It took a lot of work but possible.
Googles going to have to put down the scythe though.
To regain trust with enterprise customers (that's where the money is) it would take at least a decade of cleaning up their act very thoroughly.
I remember a poster proudly announcing that they will announce one year in advance when they plan to shut down a service on GCS.
Maybe they should look what happens when a company like HP announces end of support for OpenVMS 5 - 10 years in advance (as just an example pulled out of my ass with some Wikipedia backing the statement up) there's a lot of wailing and howling amongst their enterprise customers.
Those are the timescales big shops think in when it comes to their strategic IT decisions.
Frankly, I don't see Google ever satisfying such requirements backed by iron clad guarantees.
I disagree. It is much easier to regain trust with enterprise customers than with end users.
Enterprise costumers have contracts and lawyers on retainer. Google can draft a contract which binds them to performing a service. They can write in penalties with bite which shows to their enterprise customers that they mean it.
That is assuming that they really want to.
On the other hand this trick doesn't work with end users. They don't have lawyers to evaluate such an offer, and they have lot less of a chance of successfully suing google even if they would have favourable contracts.
Every time they take a decision like this one, they are making the hole deeper.
To be fair that's normally the case for Google.
Wifi connections having high jitter (ie. usually the link is 1ms, but sometimes spikes up to 100ms) was a big issue.
Poor software design was also a big problem - so big in fact they had to modify the android OS and Chrome to have a special low latency rendering mode that cut out chunks of the compositor. Both Windows and Mac also have issues with getting data from the network card to the screen in less than a few milliseconds.
They could have pretty much solved the latency issues with a basic predictive neural network that predicts 'given that this set of keys have been pressed, and the screen looked like this, and this number of milliseconds have passed, what will the screen look like?'. Such a neural net could be very small, since effectively it will just deal with mouse movements, panning, scrolling, and maybe flashes like gunfire. On a per-pixel basis it would simply provide motion vectors and an additive R/G/B figure.
You will own nothing and be happy. Great.